{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2025:IYEQMZRZ6TELPKECPPYIF75ZN4","short_pith_number":"pith:IYEQMZRZ","canonical_record":{"source":{"id":"2506.17211","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2025-06-20T17:59:07Z","cross_cats_sorted":[],"title_canon_sha256":"23febe0b6a82a019b23f1945be54ecbbff07490218d1e69f4cbc7bb6006bca91","abstract_canon_sha256":"d22abeac483053d1cdc9af5a8c833d325d7527fc16043a38f7c8c1c8af68f578"},"schema_version":"1.0"},"canonical_sha256":"4609066639f4c8b7a8827bf082ffb96f0ffc3ce4c50162bf4a500456165a5699","source":{"kind":"arxiv","id":"2506.17211","version":1},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2506.17211","created_at":"2026-07-05T11:24:58Z"},{"alias_kind":"arxiv_version","alias_value":"2506.17211v1","created_at":"2026-07-05T11:24:58Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2506.17211","created_at":"2026-07-05T11:24:58Z"},{"alias_kind":"pith_short_12","alias_value":"IYEQMZRZ6TEL","created_at":"2026-07-05T11:24:58Z"},{"alias_kind":"pith_short_16","alias_value":"IYEQMZRZ6TELPKEC","created_at":"2026-07-05T11:24:58Z"},{"alias_kind":"pith_short_8","alias_value":"IYEQMZRZ","created_at":"2026-07-05T11:24:58Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2025:IYEQMZRZ6TELPKECPPYIF75ZN4","target":"record","payload":{"canonical_record":{"source":{"id":"2506.17211","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2025-06-20T17:59:07Z","cross_cats_sorted":[],"title_canon_sha256":"23febe0b6a82a019b23f1945be54ecbbff07490218d1e69f4cbc7bb6006bca91","abstract_canon_sha256":"d22abeac483053d1cdc9af5a8c833d325d7527fc16043a38f7c8c1c8af68f578"},"schema_version":"1.0"},"canonical_sha256":"4609066639f4c8b7a8827bf082ffb96f0ffc3ce4c50162bf4a500456165a5699","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:24:58.303079Z","signature_b64":"LzLir/vXghCIOLJ7nfFX+befsEtgTrvN70eqDiFYJC9QdaGM/9kloYUmSF8ZseuK1WI1KVwE2Fh2Gt4wtEE0AQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"4609066639f4c8b7a8827bf082ffb96f0ffc3ce4c50162bf4a500456165a5699","last_reissued_at":"2026-07-05T11:24:58.302468Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:24:58.302468Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2506.17211","source_version":1,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T11:24:58Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"ItwkecXpn+8Hu1Hip5eue/mgppUkFCFBp/GXqjfi8Jj17qgkowOTABVXiIXFRN2GyIDgFcfJXO6Qyo/bf4iYBw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-17T01:11:33.412774Z"},"content_sha256":"866945f77e65b809a10ce115af2a13822ca8ad21bfe92994d98bf1dae67866c7","schema_version":"1.0","event_id":"sha256:866945f77e65b809a10ce115af2a13822ca8ad21bfe92994d98bf1dae67866c7"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2025:IYEQMZRZ6TELPKECPPYIF75ZN4","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"BREAD: Branched Rollouts from Expert Anchors Bridge SFT & RL for Reasoning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Chenshun Ni, Jiasi Chen, Samet Oymak, Xuechen Zhang, Yingcong Li, Zijian Huang","submitted_at":"2025-06-20T17:59:07Z","abstract_excerpt":"Small language models (SLMs) struggle to learn complex reasoning behaviors, especially when high-quality traces are scarce or difficult to learn from. The standard training approach combines a supervised fine-tuning (SFT) stage, often to distill capabilities of a larger model, followed by a reinforcement learning (RL)stage such as Group Relative Policy Optimization (GRPO). In this paper, we investigate the fundamental limitations of this SFT + RL paradigm and propose methods to overcome them. Under a suitable theoretical model, we demonstrate that the SFT + RL strategy can fail completely when"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2506.17211","kind":"arxiv","version":1},"verdict":{"id":null,"model_set":{},"created_at":null,"strongest_claim":"","one_line_summary":"","pipeline_version":null,"weakest_assumption":"","pith_extraction_headline":""},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2506.17211/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"verdict_id":null},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T11:24:58Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"1bAmogLlUMr6SYtohY4umeXuK1DUBOTh06YTXhsX92I0MfWHAHGRNl5fkKhhGwhiU1wmuUmI8hVy4ISECVEoBw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-17T01:11:33.413519Z"},"content_sha256":"bbce8a649a4ca62fca81a133713d25563ea0d3796727dfff9e49cfa65e240dc2","schema_version":"1.0","event_id":"sha256:bbce8a649a4ca62fca81a133713d25563ea0d3796727dfff9e49cfa65e240dc2"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/IYEQMZRZ6TELPKECPPYIF75ZN4/bundle.json","state_url":"https://pith.science/pith/IYEQMZRZ6TELPKECPPYIF75ZN4/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/IYEQMZRZ6TELPKECPPYIF75ZN4/bundle.json","status":"primary"}],"public_keys":[{"key_id":"pith-v1-2026-05","algorithm":"ed25519","format":"raw","public_key_b64":"stVStoiQhXFxp4s2pdzPNoqVNBMojDU/fJ2db5S3CbM=","public_key_hex":"b2d552b68890857171a78b36a5dccf368a953413288c353f7c9d9d6f94b709b3","fingerprint_sha256_b32_first128bits":"RVFV5Z2OI2J3ZUO7ERDEBCYNKS","fingerprint_sha256_hex":"8d4b5ee74e4693bcd1df2446408b0d54","rotates_at":null,"url":"https://pith.science/pith-signing-key.json","notes":"Pith uses this Ed25519 key to sign canonical record SHA-256 digests. Verify with: ed25519_verify(public_key, message=canonical_sha256_bytes, signature=base64decode(signature_b64))."}],"merge_version":"pith-open-graph-merge-v1","built_at":"2026-08-17T01:11:33Z","links":{"resolver":"https://pith.science/pith/IYEQMZRZ6TELPKECPPYIF75ZN4","bundle":"https://pith.science/pith/IYEQMZRZ6TELPKECPPYIF75ZN4/bundle.json","state":"https://pith.science/pith/IYEQMZRZ6TELPKECPPYIF75ZN4/state.json","well_known_bundle":"https://pith.science/.well-known/pith/IYEQMZRZ6TELPKECPPYIF75ZN4/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2025:IYEQMZRZ6TELPKECPPYIF75ZN4","merge_version":"pith-open-graph-merge-v1","event_count":2,"valid_event_count":2,"invalid_event_count":0,"equivocation_count":0,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"d22abeac483053d1cdc9af5a8c833d325d7527fc16043a38f7c8c1c8af68f578","cross_cats_sorted":[],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2025-06-20T17:59:07Z","title_canon_sha256":"23febe0b6a82a019b23f1945be54ecbbff07490218d1e69f4cbc7bb6006bca91"},"schema_version":"1.0","source":{"id":"2506.17211","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2506.17211","created_at":"2026-07-05T11:24:58Z"},{"alias_kind":"arxiv_version","alias_value":"2506.17211v1","created_at":"2026-07-05T11:24:58Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2506.17211","created_at":"2026-07-05T11:24:58Z"},{"alias_kind":"pith_short_12","alias_value":"IYEQMZRZ6TEL","created_at":"2026-07-05T11:24:58Z"},{"alias_kind":"pith_short_16","alias_value":"IYEQMZRZ6TELPKEC","created_at":"2026-07-05T11:24:58Z"},{"alias_kind":"pith_short_8","alias_value":"IYEQMZRZ","created_at":"2026-07-05T11:24:58Z"}],"graph_snapshots":[{"event_id":"sha256:bbce8a649a4ca62fca81a133713d25563ea0d3796727dfff9e49cfa65e240dc2","target":"graph","created_at":"2026-07-05T11:24:58Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"graph_snapshot":{"author_claims":{"count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","strong_count":0},"builder_version":"pith-number-builder-2026-05-17-v1","claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"integrity":{"available":true,"clean":true,"detectors_run":[],"endpoint":"/pith/2506.17211/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Small language models (SLMs) struggle to learn complex reasoning behaviors, especially when high-quality traces are scarce or difficult to learn from. The standard training approach combines a supervised fine-tuning (SFT) stage, often to distill capabilities of a larger model, followed by a reinforcement learning (RL)stage such as Group Relative Policy Optimization (GRPO). In this paper, we investigate the fundamental limitations of this SFT + RL paradigm and propose methods to overcome them. Under a suitable theoretical model, we demonstrate that the SFT + RL strategy can fail completely when","authors_text":"Chenshun Ni, Jiasi Chen, Samet Oymak, Xuechen Zhang, Yingcong Li, Zijian Huang","cross_cats":[],"headline":"","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2025-06-20T17:59:07Z","title":"BREAD: Branched Rollouts from Expert Anchors Bridge SFT & RL for Reasoning"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2506.17211","kind":"arxiv","version":1},"verdict":{"created_at":null,"id":null,"model_set":{},"one_line_summary":"","pipeline_version":null,"pith_extraction_headline":"","strongest_claim":"","weakest_assumption":""}},"verdict_id":null}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:866945f77e65b809a10ce115af2a13822ca8ad21bfe92994d98bf1dae67866c7","target":"record","created_at":"2026-07-05T11:24:58Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"attestation_state":"computed","canonical_record":{"metadata":{"abstract_canon_sha256":"d22abeac483053d1cdc9af5a8c833d325d7527fc16043a38f7c8c1c8af68f578","cross_cats_sorted":[],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2025-06-20T17:59:07Z","title_canon_sha256":"23febe0b6a82a019b23f1945be54ecbbff07490218d1e69f4cbc7bb6006bca91"},"schema_version":"1.0","source":{"id":"2506.17211","kind":"arxiv","version":1}},"canonical_sha256":"4609066639f4c8b7a8827bf082ffb96f0ffc3ce4c50162bf4a500456165a5699","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"4609066639f4c8b7a8827bf082ffb96f0ffc3ce4c50162bf4a500456165a5699","first_computed_at":"2026-07-05T11:24:58.302468Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T11:24:58.302468Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"LzLir/vXghCIOLJ7nfFX+befsEtgTrvN70eqDiFYJC9QdaGM/9kloYUmSF8ZseuK1WI1KVwE2Fh2Gt4wtEE0AQ==","signature_status":"signed_v1","signed_at":"2026-07-05T11:24:58.303079Z","signed_message":"canonical_sha256_bytes"},"source_id":"2506.17211","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:866945f77e65b809a10ce115af2a13822ca8ad21bfe92994d98bf1dae67866c7","sha256:bbce8a649a4ca62fca81a133713d25563ea0d3796727dfff9e49cfa65e240dc2"],"state_sha256":"243602e9f7cdfd8db4bdc918cc788fc0990a8b19ea5f967f200c845845bd9042"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"pB7xx7QrXYrPbA02GpjpeRg+OU273i7NuIrAzfQD7J2jYy/wyodhVQDXrhwFnM3vMfjYAJMAldBRtwTeUFkVAg==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-17T01:11:33.429130Z","bundle_sha256":"bea0421dafae91fd4ff39dca3a89d5b54d7d73d9afa8bb3beb166827f7e4dce0"}}